Re: [nsouch@fr.alcove.com: Re: KGI/NVIDIA/FreeBSD]

[Munish C. <mes...@ho...>]

Hiya,

Well I guess it's good I subscribed to the list. I sent Nicholas a mail with 
things that I felt were important to have NVIDIA recognize this (please 
forward that to the list, Nicholas). The main thing that would attract 
NVIDIA is write once, compile everywhere...BUT, they wouldn't like the fact 
that it isn't really usable yet as an X replacement. This is of course my 
view on things, but I can't bee too far off the mark.

One of the things I mentioned to Nicholas is that there is need for clear 
direction. As in, set up a team of people, clearly with names and functions 
and PURPOSE that wants to work on getting NVIDIA, KGI, and FreeBSD working 
together. Work out the ugly stuff BEFORE involving NVIDIA as a company.

So there are things that clearly need to be worked out. Timeframes (as in 
goals in the form of dates/performance/whatever) are important too. This 
might sound boring, but it's essential that NVIDIA sees people making an 
effort. Then come the NDA's and what not...

'Our' driver should be halfway usable by the end of the month (running the 
linux driver module in compat mode on FreeBSD), and after that we'll push 
for full native support.

Cheers,
Munish


>From: Steffen Seeger <se...@ph...>
>To: Nicolas Souchu <ns...@fr...>
>CC: KGI Devel <kgi...@li...>
>Subject: Re: [ns...@fr...: Re: KGI/NVIDIA/FreeBSD]
>Date: Fri, 11 May 2001 16:14:23 +0200
>
>Nicolas Souchu wrote:
> >
> > Folks,
> >
> > You may want to had arguments to this...
> >
> > A team is willing to provide FreeBSD with full 3D features for NVIDIA
> > boards in FreeBSD.
> >
> > A page is dedicated to this http://nvidia.netexplorer.org
>
>The main problem would be getting documentation out of NVidia.
>
> > ----- Forwarded message from Nicolas Souchu <ns...@fr...> -----
> >
> > Date: Wed, 9 May 2001 14:52:21 +0200
> > From: Nicolas Souchu <ns...@fr...>
> > To: "Pedro F. Giffuni" <pf...@pi...>
> > Cc: Munish Chopra <mes...@ho...>
> > Subject: Re: KGI/NVIDIA/FreeBSD
> > User-Agent: Mutt/1.2.4i
> > In-Reply-To: <3AF...@pi...>; from pf...@pi... on Tue, 
>May 08, 2001 at 10:17:27AM -0400
> > Organization: Alcôve, http://www.alcove.com
> >
> > On Tue, May 08, 2001 at 10:17:27AM -0400, Pedro F. Giffuni wrote:
> > > I'm CC'ing Nicolas, as he is working on FreeBSD's KGI.
> >
> > Currently, looking at the work :) Nothing's really done yet because of
> >
> >         - lack of time (like everyone :(
> >         - lack of supported HW
> >
> > Has a matter of fact, I need some well supported HW not to have to
> > bother with driver problems and focus only to the OS-dependent part.
>
>I am working with a Permedia2 or Permedia2v based board at the moment,
>so this could be considered a kind of 'reference' hardware.
>
> > Since writing drivers for KGI is theoretically strictly OS-independent,
> > any HW supported by the Linux version should do the job.
> >
> > Rodolphe is currently working on the Gx00 support on his *spare* time.
> >
> > I was thinking of a text-only port as a first cut while waiting for
> > more graphic support on the Linux side.
>
>The KGI core services should work with any VGA/MDA compatible card.
>So as long as you stay with i386 for the beginning, any card should be ok
>in the beginning.
>
> > > Munish Chopra wrote:
> > > >
> > > > >Hi Munish;
> > > > >
> > > > >I wouldn't suggest our people spending time on the KGI driver, 
>there is at
> > > > >least one person in the KGI project interested.
> > > >
> > > > I'm going to assume the 'our' was supposed to be 'your'...
> > >
> > > I mean the FreeBSD people. Jon_Taylor appears to be the maintainer for
> > > NVidia cards for KGI (Linux). I'm not sure about his address I'd have
> > > to dig it in the KGI list, perhaps Nicolas knows it?
> >
> > No, I dont.
>
>The last one I have currently at hand is "Jon M. Taylor" 
><ta...@ec...>
>
> > > > >The idea here is that if if getting a FreeBSD native driver is too
> > > > >tough...another alternative is to convince them (NVIDIA) to produce 
>a port
> > > > >for KGI/Linux. By the way KGI works, I understand it would be >very 
>simple
> > > > >to get that driver working on FreeBSD.
> >
> > It should be simple... KGI is almost autonomous and doesn't much rely on
> > Linux.
> >
> > > > I'm still not too sure, but I haven't had time to look deeper into 
>KGI
> > > > (forgive me, but the site seemed...'dead'). I hope I get some time 
>today.
> > > >
> > > > >This is, of course assuming the KGI port is successful. Just 
>between us,
> > > > >commiting KGI into FreeBSD would hit hard Linus' pride :).
> >
> > :) The most important thing would be to provide FBSD (and even later 
>*BSD) with
> > a *real* graphic solution directly reusable for boot and install ;)
>
>A view I would share. If you want to do me a favour, please do not choose
>KGI to hit Linus' pride.
>
> > XFree replacement is a challenge that should be addressed later, I 
>think.
>
>KGI is not aiming at a X replacement. It's about providing a infrastructure 
>that
>would allow X to run well, but allow for easy re-use of drivers in new 
>projects.
>In other words, decouple drivers from the windowing system, so that the 
>work
>that went into drivers can be used by other than just the X(F86) Project.
>
> > > > Oh well then there's no discussion. It must be done! :)
> > > >
> > > > Do me a favor, get me the e-mail address of the person interested in 
>doing
> > > > the KGI work. I'll take a look at things and make it somewhat 
>presentable to
> > > > Erik and the guys at fbsd-nvdriver. It's to a large extent their 
>call...but
> > > > both them and NVIDIA need to be convinced that this is useful. (if 
>you've
> > > > got any good arguments I'd like to hear them...since I still haven't 
>looked
> > > > much at KGI I'm relatively blank...)
> >
> > Arguments? For the KGI specific aspects, you should ask S.Seeger for one 
>or
> > two of his mails about the subject. He certainly has in mind more 
>arguments
> > than we can imagine. Just start a thread arguing on XFree and 
>framebuffer
> > features on kgi-develop and you'll see ;)
>
>Well, the main argument in favour I have would be that I have tried very 
>hard to
>separate the stuff so that
>
>a) the drivers should work with at most a recompiliation
>	on all platforms that have the KGI core ported,
>b) the main driver functionality is independent of the application
>	and it's hardware model (e.g. mode checking, not enumeration)
>c) the driver architecture is modularize as to maximize reuse of common
>	parts as much as possible.
>
>Of course, I don't want to keep the arguments against it away from you:
>
>a)	it still needs work to get to a useable state to be a potential XFree86
>	replacement for production purposes.
>b)	I cannot commit full-time to it's development. I need to finish my
>	PhD this year, and I don't know yet what's coming afterwards, so I can't
>	make promises now.
>c)	It has had much controverse discussions in the past.
>
> > For the FreeBSD side, well... FBSD is already a very good OS for many 
>things
> > other than graphics. KGI is technically
> > very good but is not considered as it should by the Linux camp. The port
> > IS the solution to address both issues at the same. But you certainly 
>need
> > more than this as a proposal to NVIDIA. The good arguments should come 
>from
> > Seeger as he is the leader of the project.
>
>At least it would be a good point for graphics vendors to consider a way to
>have drivers on both Linux and FreeBSD with one development effort.
>Which is the main point why KGI started. We want the hardware companies to 
>be
>able to write one driver, which can be used on several platforms and by 
>several
>user interface projects, not just X11.
>
>I don't think a useable X11 alternative will be available soon, nor do I 
>think
>it would be an easy task to tackle, but I don't want it to become 
>impossible
>because
>it will not have any useable driver base to work on, just because it is
>different
>than X. So, first objective for the moment is a 'better (== more modular)' 
>X
>driver infrastructure that __allows__ other projects to recycle the driver 
>work.
>
> > As far as I could see, the *BSD world and especially the FreeBSD team 
>would
> > be really pleased to include KGI in there project (thanks to Seeger's 
>license).
>
>As for the core stuff most of it is BSD style or GPL, as you like it.
>
> > If things become more serious (especially if NVIDIA is interested by 
>KGI),
> > I could ask my employer for more time for the port (as I did it in the 
>past
> > for the GGI port).
>
>If NVidia becomes interested in KGI I should probably think of getting 
>hired
>by them (or at least get some support/financial motiviation, whatever you 
>call
>it
>:-)))
>
> > > KGI is great, Nicolas can give more arguments :)... the nice thing is
> > > that we would get all the graphic applications accelerated (including
> > > Linux emulated stuff), not just XWindows. The problem I see now is
> > > that although Nicolas is very capable, I suspect the port won't be
> > > ready before 5.0 Release (which I've heard will be out by the end of
> > > this summer!).
> >
> > Even if it was included in the 5.0, it would only be experimental, so
> > it has nothing to do in a -stable branch.
>
>I am quite open to a BSD port, but I cannot work on it on my own.
>I have digged through the Linux Kernel, through the X11 sample 
>implementation,
>the OpenGL sample implementation. I think I can't cope with *BSD in 
>addition
>to that (though I will assist with porting stuff, technical questions, 
>etc.)
>
>Yours,
>			Steffen
>
>_______________________________________________________________________________
>Steffen Seeger                              
>mailto:se...@ph...
>TU-Chemnitz                                      
>http://www.tu-chemnitz.de/~sse
>
>_______________________________________________
>kgi-develop mailing list
>kgi...@li...
>http://lists.sourceforge.net/lists/listinfo/kgi-develop

_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.